MoEngage Converts 400 Private APIs Into Agent Tools Using AgentCore Gateway
The engagement platform used Amazon Bedrock AgentCore to expose internal REST endpoints safely, avoiding public internet exposure and per-service server overhead.
Key points
- MoEngage converted roughly 400 private REST APIs into permission-scoped agent tools.
- The solution uses Amazon Bedrock AgentCore Gateway across seven AWS Regions.
- No public endpoints or per-service Model Context Protocol servers were created.
MoEngage has integrated approximately 400 internal REST APIs as agent tools via Amazon Bedrock AgentCore Gateway, according to an advisory from AWS Machine Learning.reported This implementation spans seven AWS Regions and keeps all endpoints within private virtual private clouds, avoiding public internet exposure entirely.
Evidence: reported
How MoEngage turned over 400 APIs into agent tools with AgentCore Gatewayaifleet-aws
The announcement
The customer engagement platform, used by more than 1,350 consumer brands, faced challenges in making its internal APIs safely reachable by AI agents.reported Behind its product sit roughly 400 APIs built by 25 engineering teams.reported These systems run in isolated VPCs to meet data residency commitments and are not exposed publicly. The company needed a way to grant agents access without flooding context windows with hundreds of tool definitions or giving models control over identity.
Evidence: reported
How MoEngage turned over 400 APIs into agent tools with AgentCore Gatewayaifleet-aws
Evidence: reported
How MoEngage turned over 400 APIs into agent tools with AgentCore Gatewayaifleet-aws
The context
MoEngage rejected the standard approach of having each engineering team build its own Model Context Protocol server. The company noted that over 20 such servers would create significant monitoring, patching and compatibility burdens.reported Cross-domain agents would struggle to connect to multiple servers and reconcile data. Furthermore, requiring every team to acquire MCP expertise would distract from their core product work.
Evidence: reported
How MoEngage turned over 400 APIs into agent tools with AgentCore Gatewayaifleet-aws
What product readers should know
Instead of decentralised MCP implementation, MoEngage established that individual teams only build REST APIs and declare operations as tools via OpenAPI specs. The AI Platform team owns the MCP layer end to end. This centralised approach uses AgentCore Gateway to create a single tool surface for agents. It enforces per-agent authorisation server-side using Cedar policies and managed Amazon VPC Lattice egress, eliminating the need for per-integration networking provisions.
Why it matters
- Engineering teams can avoid the operational overhead of maintaining dozens of separate MCP servers.
- Private APIs can be exposed to agents without creating public endpoints or complex networking tunnels.
- Onboarding new APIs reportedly takes hours rather than days, saving estimated months of platform engineering effort.
- Context window bloat is mitigated by centralising tool discovery instead of exposing hundreds of definitions directly.
Topics: Amazon Bedrock, AI agents, API integration, MoEngage, AgentCore Gateway
Updated
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Questions readers ask
Does this solution require exposing internal APIs to the public internet?
No, the architecture keeps all internal APIs within private VPCs with no public endpoints in the path.
Who is responsible for implementing the Model Context Protocol layer?
The AI Platform team owns the MCP layer end to end, while individual engineering teams only provide OpenAPI specs.
How many APIs were converted into agent tools?
MoEngage turned roughly 400 internal REST APIs into permission-scoped agent tools.
Connected stories
Reading receipt
- MoEngage converted roughly 400 private REST APIs into permission-scoped agent tools.
- The solution uses Amazon Bedrock AgentCore Gateway across seven AWS Regions.
- No public endpoints or per-service Model Context Protocol servers were created.